Hi, Elisabeth, from this site:
http://github.com/dirkneumann/spm8/blob/master/spm_eeg_morlet.m
it looks like the extent of the Morlet window is about 0.0071 seconds
(full-width half-max).
In the frequency domain this window will extend about 140 Hz.
The complex Morlet wavelet coefficient, at a specific time and frequency,
is the discrete Fourier transform of Gaussian windowed data (centered on
the time of interest) evaluated at the frequency of interest. This is
equivalent in the frequency domain to the convolution of the transformed
Gaussian window with the transformed data (evaluated at the frequency of
interest).
Synopsis: with a scale factor of 7, estimates at 40-80 Hz will respond to
frequencies from approximately Nyquist - 30 Hz to 150 Hz. If this is the
problem, the solution will be to increase the scale factor; trading
temporal resolution for resolution in the frequency domain.
Though possibly pre-processed away or taken care of, the power lines
belong to these frequency ranges. If you haven't already done so, it's
probably a good idea to make sure that they are not contributing.
Best,
Kyle Lepage